OTiS Pre-training Corpus

该数据集是一个庞大且多样化的公开时间序列数据集，涵盖了8个领域。它包含了总共640,187个样本和110亿个时间点。该数据集支持多种任务，包括二元癫痫检测、多类别故障检测、手势分类、肌肉疾病分类，以及变压器温度、天气和电力消耗的预测。其规模之大，样本数量之丰富，为时间序列分析中的分类、回归和预测等多种任务提供了强有力的支持。

This is a large-scale and diverse public time series dataset covering 8 domains. It contains a total of 640,187 samples and 11 billion time points. This dataset supports a variety of tasks, including binary epilepsy detection, multi-class fault detection, gesture classification, muscular disease classification, as well as forecasting of transformer temperature, weather and power consumption. With its large scale and rich sample volume, it provides robust support for multiple tasks in time series analysis such as classification, regression and forecasting.

搜集汇总

背景与挑战

背景概述

OTiS预训练语料库是一个用于通用时间序列分析的数据集，支持多领域时间序列任务。该数据集通过预训练模型权重和标准化数据预处理流程，旨在提升时间序列模型的泛化能力，适用于分类、回归和预测等多种下游任务。